Abstract: Tuberculosis (TB) is the world's leading infectious killer with 1.8 million deaths in 2015 as reported by WHO. It is therefore imperative that alternate routes of identification of novel anti-TB compounds are explored given the time and costs involved in new drug discovery process. Towards this, we have developed RepTB. This is a unique drug repurposing approach for TB that uses molecular function correlations among known drug-target pairs to predict novel drug-target interactions. In this study, we have created a Gene Ontology based network containing 26,404 edges, 6630 drug and 4083 target nodes. The network, enriched with molecular function ontology, was analyzed using Network Based Inference (NBI). The association scores computed from NBI are used to identify novel drug-target interactions. These interactions are further evaluated based on a combined evidence approach for identification of potential drug repurposing candidates. In this approach, targets which have no known variation in clinical isolates, no human homologs, and are essential for Mtb's survival and or virulence are prioritized. We analyzed predicted DTIs to identify target pairs whose predicted drugs may have synergistic bactericidal effect. From the list of predicted DTIs from RepTB, four TB targets, namely, FolP1 (Dihydropteroate synthase), Tmk (Thymidylate kinase), Dut (Deoxyuridine 5'-triphosphate nucleotidohydrolase) and MenB (1,4-dihydroxy-2-naphthoyl-CoA synthase) may be selected for further validation. In addition, we observed that in some cases there is significant chemical structure similarity between predicted and reported drugs of prioritized targets, lending credence to our approach. We also report new chemical space for prioritized targets that may be tested further. We believe that with increasing drug-target interaction dataset RepTB will be able to offer better predictive value and is amenable for identification of drug-repurposing candidates for other disease indications too.

Abstract: BACKGROUND: The right drug to the right patient at the right time is one of the ideals of Individualized Medicine (IM) and remains one of the most compelling promises of the post-genomic age. The addition of genomic information is expected to increase the precision of an individual patient's treatment, resulting in improved outcomes. While pilot studies have been encouraging, key aspects of interpreting tumor genomics information, such as somatic activation of drug transport or metabolism, have not been systematically evaluated. METHODS: In this work, we developed a simple rule-based approach to classify the therapies administered to each patient from The Cancer Genome Atlas PanCancer dataset (n = 2858) as effective or ineffective. Our Therapy Efficacy model used each patient's drug target and pharmacokinetic (PK) gene expression profile; the specific genes considered for each patient depended on the therapies they received. Patients who received predictably ineffective therapies were considered at high-risk of cancer-related mortality and those who did not receive ineffective therapies were considered at low-risk. The utility of our Therapy Efficacy model was assessed using per-cancer and pan-cancer differential survival. RESULTS: Our simple rule-based Therapy Efficacy model classified 143 (5%) patients as high-risk. High-risk patients had age ranges comparable to low-risk patients of the same cancer type and tended to be later stage and higher grade (odds ratios of 1.6 and 1.4, respectively). A significant pan-cancer association was identified between predictions of our Therapy Efficacy model and poorer overall survival (hazard ratio, HR = 1.47, p = 6.3 x 10(- 3)). Individually, drug export (HR = 1.49, p = 4.70 x 10(- 3)) and drug metabolism (HR = 1.73, p = 9.30 x 10(- 5)) genes demonstrated significant survival associations. Survival associations for target gene expression are mechanism-dependent. Similar results were observed for event-free survival. CONCLUSIONS: While the resolution of clinical information within the dataset is not ideal, and modeling the relative contribution of each gene to the activity of each therapy remains a challenge, our approach demonstrates that somatic PK alterations should be integrated into the interpretation of somatic transcriptomic profiles as they likely have a significant impact on the survival of specific patients. We believe that this approach will aid the prospective design of personalized therapeutic strategies.

Abstract: BACKGROUND: High throughput sequencing technologies have been an increasingly critical aspect of precision medicine owing to a better identification of disease targets, which contributes to improved health care cost and clinical outcomes. In particular, disease-oriented targeted enrichment sequencing is becoming a widely-accepted application for diagnostic purposes, which can interrogate known diagnostic variants as well as identify novel biomarkers from panels of entire human coding exome or disease-associated genes. RESULTS: We introduce a workflow named VAReporter to facilitate the management of variant assessment in disease-targeted sequencing, the identification of pathogenic variants, the interpretation of biological effects and the prioritization of clinically actionable targets. State-of-art algorithms that account for mutation phenotypes are used to rank the importance of mutated genes through visual analytic strategies. We established an extensive annotation source by integrating a wide variety of biomedical databases and followed the American College of Medical Genetics and Genomics (ACMG) guidelines for interpretation and reporting of sequence variations. CONCLUSIONS: In summary, VAReporter is the first web server designed to provide a "one-stop" resource for individual's diagnosis and large-scale cohort studies, and is freely available at http://rnd.cgu.edu.tw/vareporter .

Abstract: The receptor tyrosine kinase, erythropoietin-producing hepatocellular A4 (EphA4), was recently identified as a molecular target for Alzheimer's disease (AD). We found that blockade of the interaction of the receptor and its ligands, ephrins, alleviates the disease phenotype in an AD transgenic mouse model, suggesting that targeting EphA4 is a potential approach for developing AD interventions. In this study, we identified five FDA-approved drugs-ergoloid, cyproheptadine, nilotinib, abiraterone, and retapamulin-as potential inhibitors of EphA4 by using an integrated approach combining virtual screening with biochemical and cellular assays. We initially screened a database of FDA-approved drugs using molecular docking against the ligand-binding domain of EphA4. Then, we selected 22 candidate drugs and examined their inhibitory activity towards EphA4. Among them, five drugs inhibited EphA4 clustering induced by ephrin-A in cultured primary neurons. Specifically, nilotinib, a kinase inhibitor, inhibited the binding of EphA4 and ephrin-A at micromolar scale in a dosage-dependent manner. Furthermore, nilotinib inhibited the activation of EphA4 and EphA4-dependent growth cone collapse in cultured hippocampal neurons, demonstrating that the drug exhibits EphA4 inhibitory activity in cellular context. As demonstrated in our combined computational and experimental approaches, repurposing of FDA-approved drugs to inhibit EphA4 may provide an alternative fast-track approach for identifying and developing new treatments for AD.

Abstract: There are many mechanisms of neurotoxicity that are initiated by the interaction of chemicals with different neurological targets. Under the U.S. Environmental Protection Agency's ToxCast program, the biological activity of thousands of chemicals was screened in biochemical and cell-based assays in a high-throughput manner. Two hundred sixteen assays in the ToxCast screening database were identified as targeting a total of 123 proteins having neurological functions according to the Gene Ontology database. Data from these assays were imported into the Organization for Economic Co-operation and Development QSAR Toolbox and used to predict neurological targets for chemical neurotoxins. Two sets of data were generated: one set was used to classify compounds as active or inactive and another set, composed of AC50s for only active compounds, was used to predict AC50 values for unknown chemicals. Chemical grouping and read-across within the QSAR Toolbox were used to identify neurologic targets and predict interactions for pyrethroids, a class of compounds known to elicit neurotoxic effects in humans. The classification prediction results showed 79% accuracy while AC50 predictions demonstrated mixed accuracy compared with the ToxCast screening data.

Abstract: Asthma is a chronic disease that is caused by airway inflammation. The main features of asthma are airway hyperresponsiveness (AHR) and reversible airway obstruction. The disease is mainly managed using drug therapy. The current asthma drug treatments are divided into two categories, namely, anti-inflammatory drugs and bronchodilators. However, disease control in asthma patients is not very efficient because the pathogenesis of asthma is complicated, inducing factors that are varied, such as the differences between individual patients. In this paper, we delineate the genetic mechanisms of asthma, and present asthma-susceptible genes and genetic pharmacology in an attempt to find a diagnosis, early prevention, and treatment methods for asthma. Finally, we reposition some clinical drugs for asthma therapy, based on asthma genetics.

Abstract: Drug interactions, including drug-drug interactions (DDIs) and drug-food constituent interactions (DFIs), can trigger unexpected pharmacological effects, including adverse drug events (ADEs), with causal mechanisms often unknown. Several computational methods have been developed to better understand drug interactions, especially for DDIs. However, these methods do not provide sufficient details beyond the chance of DDI occurrence, or require detailed drug information often unavailable for DDI prediction. Here, we report development of a computational framework DeepDDI that uses names of drug-drug or drug-food constituent pairs and their structural information as inputs to accurately generate 86 important DDI types as outputs of human-readable sentences. DeepDDI uses deep neural network with its optimized prediction performance and predicts 86 DDI types with a mean accuracy of 92.4% using the DrugBank gold standard DDI dataset covering 192,284 DDIs contributed by 191,878 drug pairs. DeepDDI is used to suggest potential causal mechanisms for the reported ADEs of 9,284 drug pairs, and also predict alternative drug candidates for 62,707 drug pairs having negative health effects. Furthermore, DeepDDI is applied to 3,288,157 drug-food constituent pairs (2,159 approved drugs and 1,523 well-characterized food constituents) to predict DFIs. The effects of 256 food constituents on pharmacological effects of interacting drugs and bioactivities of 149 food constituents are predicted. These results suggest that DeepDDI can provide important information on drug prescription and even dietary suggestions while taking certain drugs and also guidelines during drug development.

Abstract: Development of targeted nanoparticle drug carriers often requires complex synthetic schemes involving both supramolecular self-assembly and chemical modification. These processes are generally difficult to predict, execute, and control. We describe herein a targeted drug delivery system that is accurately and quantitatively predicted to self-assemble into nanoparticles based on the molecular structures of precursor molecules, which are the drugs themselves. The drugs assemble with the aid of sulfated indocyanines into particles with ultrahigh drug loadings of up to 90%. We devised quantitative structure-nanoparticle assembly prediction (QSNAP) models to identify and validate electrotopological molecular descriptors as highly predictive indicators of nano-assembly and nanoparticle size. The resulting nanoparticles selectively targeted kinase inhibitors to caveolin-1-expressing human colon cancer and autochthonous liver cancer models to yield striking therapeutic effects while avoiding pERK inhibition in healthy skin. This finding enables the computational design of nanomedicines based on quantitative models for drug payload selection.

Abstract: Identification of autoimmune processes and introduction of new autoantigens involved in the pathogenesis of multiple sclerosis (MS) can be helpful in the design of new drugs to prevent unresponsiveness and side effects in patients. To find significant changes, we evaluated the autoantibody repertoires in newly diagnosed relapsing-remitting MS patients (NDP) and those receiving disease-modifying therapy (RP). Through a random peptide phage library, a panel of NDP- and RP-specific peptides was identified, producing two protein data sets visualized using Gephi, based on protein--protein interactions in the STRING database. The top modules of NDP and RP networks were assessed using Enrichr. Based on the findings, a set of proteins, including ATP binding cassette subfamily C member 1 (ABCC1), neurogenic locus notch homologue protein 1 (NOTCH1), hepatocyte growth factor receptor (MET), RAF proto-oncogene serine/threonine-protein kinase (RAF1) and proto-oncogene vav (VAV1) was found in NDP and was involved in over-represented terms correlated with cell-mediated immunity and cancer. In contrast, transcription factor RelB (RELB), histone acetyltransferase p300 (EP300), acetyl-CoA carboxylase 2 (ACACB), adiponectin (ADIPOQ) and phosphoenolpyruvate carboxykinase 2 mitochondrial (PCK2) had major contributions to viral infections and lipid metabolism as significant events in RP. According to these findings, further research is required to demonstrate the pathogenic roles of such proteins and autoantibodies targeting them in MS and to develop therapeutic agents which can ameliorate disease severity.

Abstract: A variety of computational screening methods generate similarity-based compound rankings for hit identification. However, these rankings are difficult to interpret. It is essentially impossible to determine where novel active compounds might be found in database rankings. Thus, compound selection largely depends on intuition and guesswork. Herein, we show that molecular networks can substantially aid in the analysis of similarity-based compound rankings. A series of networks generated for rankings provides visual access to search results and adds chemical neighborhood and context information for reference compounds that are not available in rankings. Network structure is shown to serve as a diagnostic criterion for the likelihood to successfully select active compounds from rankings. In addition, comparison of different networks makes it possible to prioritize alternative similarity measures for search calculations and optimize the enrichment of active compounds in rankings.